AILab Howest

Use case

VEG-i-TEC has awhitenessThis is a machine for briefly and effectively steaming fries and vegetables. Hot wastewater flows from this machine to the sewer, but there is still a lot of thermal energy in it.

Our research focuses on recovering this energy: we want to transfer the heat from the wastewater to fresh water and then store it in abuffer tankThis preheated water can then be used in other processes that require hot water, instead of having to heat cold tap water. Depending on the user, this leads to a reduction in electricity consumption or gas consumption.

Challenges

Before a company invests in a buffer tank system, it is essential to knowhow much gas or electricity can potentially be saved, and what theoptimal capacityof the tank would be.

Industrial processes are often very complex and dynamic, making these estimates very difficult. Therefore, we developed a user-friendly simulation tool to simplify this process.

Simulation tool

After logging into the application, the user sees an overview of all their projects. Each project consists of a block diagram, where each block represents a process component (machine, tank, …). The connections between blocks represent the pipes through which liquids are pumped.

In the left sidebar, there are a number ofpre-made components, but you can also create a new one yourself. The intention is that one can later manage their nodes.publishso that other people can also use these to create a largelibraryto build.

On the right side, one can find allpropertiesadjust the selected component or tube. The most notable feature is the Python code section. The user can actually thebehaviorfrom machinesself-programmingwith theCoolProplibrary (physical calculations). This can of course be done manually, but there is also aAI buttonprovided to have the code generated by ChatGPT. You can therefore consider this schema as achain of Python functions(pieces of code) that are linked sequentially.

If we want to simulate 100 hours with an interval of 1 minute, the simulation will consist of 6000 steps. This means we execute the chain of Python functions once and repeat this 6000 times. After each step, we keep track of the temperatures, pressures, flow rates, and thermal energies of all fluids.

Technical elaboration

The website is built withNext.jsa modernReact frameworkthat server-side rendering is supported. For the backend, we useSupabasean open-source alternative forFirebasethat provides a solution for:

• Database management (PostgreSQL)
• Authentication
• Inventory storage
• Real-time data
• REST and GraphQL APIs

Via Supabase it is only possible to create JavaScript APIs, but since the simulations use Python code, this is not suitable. Technically, there are libraries such asPyodidethat make it possible to execute Python code in a JavaScript environment. We chose to develop a separate API in Python for now withFastAPI.